krishna prasad vadrevu and lee ellenberg nasa …...to further correct for scene-to-scene...

Atmospheric Correction of Sentinel 2A Data and Influence on Winter Wheat Phenology in Kabul

Region, Afghanisthan

Krishna Prasad Vadrevu and Lee Ellenberg NASA MSFC

https://ntrs.nasa.gov/search.jsp?R=20170008841 2020-08-06T08:30:06+00:00Z

• Is Atmospheric correction needed?

• How much is the improvement in scene qualitybefore and after atmospheric correction (specific toKabul region in Afghanistan for NDVI)?

• Is the improvement consistent across all months?

Atmospheric Correction of Sentinel-2a Data

Units of Electromagnetic Radiation

• Irradiance - radiant flux incident on areceiving surface from all directions, Energy per unit surface area, W m-2

• Radiance - radiant flux emitted orscattered by a unit area of surface asmeasured through a solid angle, W m-2 sr-

1 µm-1(energy (Watt) per unit area (squaremeter) per solid angle per unitwavelength (µm-1))

• Reflectance - fraction of the incident fluxthat is reflected by a medium

Basic Pre-processing

Digital Number

At Sensor Radiance

Top of Atmosphere (Reflectance)

Surface Reflectance

Conversion to Radiance

Solar correction /Thermal

calibration

Atmospheric Correction

•DN to Radiance Conversion

•L = [(LMAX - LMIN)/255] x DN + LMIN

•Where LMAX = radiance at which channel saturates LMIN = minimum recordable radiance

Digital Number to Radiance

To further correct for scene-to-scene differences in solar illumination, it is useful to convert toat-satellite reflectance. The term “at-satellite” refers to the fact that this conversion does notaccount for atmospheric influences.

At-Satellite Reflectance, pλ = (π Lλ d2 ) / (ESUNλ cosθ )

Where

Lλ = spectral radiance measured for the specific waveband

θ = solar zenith angle

ESUN = mean solar exoatmospheric irradiance (W m-2 um-1), specificto the particular wavelength interval for each waveband, consult the sensor documentation.

d = Earth-sun distance in astronomical units, ranges from approx.0.9832 to 1.0167, consult an astronomical handbook for the earth-sun distance for the imagery acquisition date

At-Satellite Reflectance

• Conversion from DN to reflectance can be affected by:• Illumination geometry• Sensor calibration• Atmospheric conditions

• Since Multi-temporal images are acquired• during different times (eg: 1st day of summer longest day of sunlight)• atmospheric conditions• solar illumination• sensor calibration• view angles

Atmospheric correction is required.

Atmospheric Correction is Mandatory When Using Multi-Temporal Images

Atmospheric Layers and Constituents

Major subdivisions of the atmosphere and the types of molecules and aerosols found in each layer.

Diameter of particles less than thewavelength of radiation

Diameter of particles larger thanthe wavelength of radiation

Both Rayleigh and Mie scattering causes Haze and decreases Contrast !

Absorption is dominant process in visible Scattering is dominant process in near infraredWater absorption is increasingly important with increasing wavelength in the infrared.

Atmospheric Windows

Atmospheric correction algorithms should take into account absorption as well as scattering in different wavelengths.

Atmospheric Correction Methods

Absolute-Methods account for view angles, atmospheric

conditions, aerosol types, water vapor, gaseous absorption,etc.

-Generally Absolute methods use some sort of radiative transfer code such as MODTRAN (MODerateresolution atmospheric TRANsmission code (developed by Air Force Resarch Lab) to model atmospheric propagation ofelectromagnetic radiation (eg: 6S, FLAASH or, ATCOR)

-For Absolute methods, atmosphericproperties as inputs are required!

Relative• Relative methods all proceed under the assumption that the relationship

between the TOA radiances recorded at two different times from regions ofconstant reflectance is spatially homogeneous and can be approximated by a linear function.

• The normalization process can then be reduced to a linear regression calculation for each spectral band.

-Histogram matching-Image regression

• The main difficulty of relative normalization methods isdetermining the landscape features whose reflectance’sare nearly constant over time.

• Invariant Targets – eg: Buildings – similar solar illuminationconditions and phenological conditions

Atmospheric Correction Methods

• Sentinel-2A was launched on June 23, 2015.

• The atmospheric correction software Sen2Cor was implemented by Telespazio France and DLR on behalf of ESA. TPZ-F and DLR have teamed up in order to provide the calibration and validation of the Level-2A processor Sen2Cor.

• Sen2Cor can be obtained downloading the S2 Toolbox (http://step.esa.int/main/download) and following plugins installation procedure

Sen2Cor - Sentinel-2a Data

• Input: Level-1C ortho-image Top-Of-Atmosphere (TOA) reflectance products

• Output (60m, 20m, 10m)• Bottom-Of-Atmosphere (BOA) corrected reflectance• Aerosol Optical Thickness (AOT) map• Water Vapour (WV) map• Scene Classification (SC) map (cloud, dark areas, bare soil,

vegetation, cloud probabilities, snow, etc.)• Quality Indicators for cloud and snow probabilities

Sen2Cor – Inputs/Outputs

Afghanisthan, Kabul

Kabul Region Overlay over Sentinel-2a Data

2015_Nov_29 1

2015_Dec_01 20.5

2015_Dec_29

32016_Jan_15 40.2

2016_Feb_15 50.1

2016_Mar_15 6

2016_Mar_31 7

2016_Apr_14 8Uncor_Max

2016_Apr_27 9

0

0.4

0.3

0.6

0.7

NDV

I

Winter Wheat Phenology - Kabul

Only part of the Wheat Growing Season images could be used due to too bad image quality !

2015_12_01 - YES 2015_12_16 - NO 2015_12_29 - YES

2016_01_15 - YES 2016_03_28 - NO2016_02_15 - YES

2016_03_31 - YES 2016_04_14 2016_04_27

2016_03_31RGB

2016_03_31RGB!

2015_12_29 - NDVI

10 5 0 10 Miles

NOVIValue

- - High : 1

m llllL Low : O

Clouds -Before and After Correction

2015-12-01 – Sentinel2- L1C (RGB)

2015-12-01 – Sentinel2- L2C (RGB)

Haze-Before and After Correction

Haze before correction; 2015-12-01 – Sentinel2- L1A (B1)

Haze removal after correction; 2015-12-01 – Sentinel2- L2C (B1)

NDVI -Before and After Correction

2016_04_27NDVI -Before

2016_04_27NDVI –After

False Color Composite -Sentinel- 4,3,2

L1C – 2015_11_29

L2A – 2015_11_29

PlantingOctober-November

HarvestJanuary-February

• Is Atmospheric correction needed?

• How much is the improvement in scene qualitybefore and after atmospheric correction?

• Is the improvement consistent across all months?

Questions - Revisited

0.8

0.7

0.6

0.5

0.4

0.3

0.2

0.1

0

NDV

I

Uncor_Max Corr_Max

Results – Uncorrected vs. Corrected NDVI for Wheat

24.69

30.23

31.81

23.24

24.28

33.33

0 5

% increase10 15

Corr_Max20 25

Uncor_Max30 35

2015_12_01

2015_12_29

2016_01_15

2016_02_15

2016_03_15

2016_03_31

Percent Increase In Surface Reflectance

• Across all images, on an average 27.93%improvement has been observed after correction.

• The improvement ranged from 23.24% - 33.33%.

• It is worth using surface reflectance images ratherthan top of reflectance images (over the Kabul region in Afghanistan)

• Next Step – compare 6S with Sen2Cor

Results

Land Cover/Land Use Change SARI International Regional Science Meeting in South/Southeast Asia, Chiang Mai, Thailand 17-19th, 2017

The exercise aims to bring together available AC processors (actually 14 processors including SEN2COR, MACCS, L8-S2-6SAC, …) to generate the corresponding SR products.

The input data will be Landsat-8 and Sentinel-2 imagery of various test sites, i.e. coastal, agricultural, forest, snow/artic areas and deserts.

Objectives• To better understand uncertainties and issues on L8 and S2 AC

products• To propose further improvements of the future AC schemes

* 1st Workshop in June 21st-22nd 2016 @ University of Maryland: to elaborate concepts, protocolsand guidelines for the inter-comparison and validation of SR products• 2nd workshop in April 2017• Results to be finalized in fall 2017

https://earth.esa.int/web/sppa/meetings-workshops/acix

ACIX: CEOS-WGCV Atmospheric Correction Inter-comparison Exercise (ESA/NASA/UMD)

Demonstration of tool with much more specifics (on atmospheric parameterization file) will be presented after Day-2 - Krishna